SPMSM drives are generally of low power category and
require simple and cost-effective controllers. This paper
investigates a robust control of a position sensorless surface
permanent magnet synchronous motor drive. A mathematical
model of SPMSM using the flux estimator is utilized to
estimate the rotor position. A neural network based adaptive
integration methodology is proposed for stator flux
estimation to improve the transient and steady state
performances. Estimated flux is obtained from speed error
and current by an integrating method accomplished by
Programmable Cascaded Low Pass Filter (PCLPF)
implemented by neural network. The neural network consists
of recurrent neural network and a Feed-Forward Artificial
Neural Network (FFANN). Moreover, the hysteresis
controller is used to control the current in such a way that it
can follow the command current as close as possible to the
sinusoidal reference. The performances of the proposed drive
system have also been studied due to load torque change,
parameter variations, and speed reversal.